[crrf]inject pd(leader)_to_pd(follower-all)_network_partition, downstream s3 sync task fail

## Bug Report

Please answer these questions before submitting your issue. Thanks!

### 1. Minimal reproduce step (Required)

1. create cluster and enable crrf
2. inject pd(leader)_to_pd(follower-all)_network_partition
3. check upstream and downstream checkpoint

### 2. What did you expect to see? (Required)
pd leader handover
upstream and downstream checkpoint can proceed after the fault is recovered.

### 3. What did you see instead (Required)
downstream task is fail

### 4. What is your TiDB version? (Required)